CDX Manager

Stop guessing which AI account still has quota. cdx tracks the rate-limit window of every Codex and Claude account you own, tells you which one is usable right now, and launches it with isolated auth — in one command.

If you pay for several AI coding subscriptions, you know the dance: you hit a 5-hour limit mid-task, guess which other account still has room, re-authenticate, and lose your context on the way. cdx turns that into cdx next.

Quick Start

npm install -g cdx-manager

cdx add codex work      # register an isolated Codex profile named "work"
cdx add claude perso    # and a Claude one
cdx status              # every account's remaining quota, side by side
cdx next                # which account should I use right now?

Supported providers: codex, claude, antigravity, ollama.

cdx next answers the question you actually have, and prints the command to run:

The Core Five

Everything else is optional. These five cover the daily loop:

Table of Contents

• Quick Start
• The Core Five
• What it does
• Technical Overview
• Getting Started
• All Commands
• JSON Output
• Available Scripts
• Windows Support
• Project Structure
• Data Layout
• Troubleshooting
• Contributing
• License

What it does

Quota-aware routing

The reason cdx exists: knowing, across every account you own, which one you can actually use.

• Usage at a glance. cdx status shows token usage, 5-hour window quota, weekly quota, last-updated timestamps, priority guidance, and the last launched session in one aligned table.
• Pick for me. cdx next selects the best available assistant from your pinned priorities and the live quota picture, shows the reason, and prints the exact command to run.
• Agent notifications. When a session finishes a turn or blocks waiting for you, cdx raises a desktop notification naming the session and the repository — so you can run several in parallel without watching any of them. Opt in per session with cdx set <name> --notify on.
• Next-ready notification. cdx ready schedules a native system notification for the next assistant that comes back from cooldown, then returns immediately.
• Passive status resolution. Codex status is read from the local Codex app-server rate-limit API when available, with legacy transcript/history parsing kept as a fallback.
• Banked reset visibility. Eligible Codex accounts show the number of manually redeemable bonus resets in cdx status and expose reset details in JSON output.

Isolated multi-account sessions

• Multiple providers, one tool. Register as many Codex, Claude, Antigravity, or Ollama sessions as you need. Codex and Claude get isolated auth environments; Antigravity is launchable through agy with OS-keyring auth; Ollama runs local models through ollama run.
• Instant launch. cdx work opens your "work" session. cdx personal opens another. No config files to edit mid-flow.
• Quick relaunch. cdx last reopens the most recently launched assistant profile.
• Resume by identity, not by recency. Each session records the provider's own conversation id — imposed at launch for Claude, read back from the rollout for Codex — so cdx resume <name> reopens that conversation regardless of your working directory or of which session ran last. cdx can-resume --json reports the strategy and where the id came from, so an automated caller can tell a named resume from a best-effort one.
• Auth guardrails. cdx checks authentication before launching. If a session is not logged in, it tells you exactly what to run — no silent failures.
• Persistent launch settings. Pin per-session power, permission, and fast-mode preferences once; cdx reapplies them on every launch until you unset them.
• Spend ceiling and model fallback. cdx set <name> --budget 5 caps what a headless run may spend, and --fallback-model sonnet,haiku lets a run continue on the next model when the first is overloaded. Both are Claude Code flags that the provider accepts with --print only, so they apply to cdx run and not to interactive launches; cdx configs says so.
• Session control. Disable a session without deleting it when an account is temporarily out of credits; disabled sessions remain visible and sort last.
• Optional session labels. cdx label <name> <label> adds a short human label; list and full status tables show a LABEL column only when at least one session has one.
• Clean removal. cdx rmv wipes a session and its entire auth directory. No orphaned files, no stale credentials.

Headless automation

• One task, one JSON result. cdx run executes a prompt against a chosen or auto-selected session and returns a stable payload; --detach returns a run_id immediately.
• Survives a rate limit. cdx run --failover notices that a run stopped because its account ran out, moves the task to the next account the ranking allows, and carries on — one run_id for the whole task, with every account it occupied listed in cdx run-report. A spent account is always confirmed against its own refreshed quota first, so a healthy account is never abandoned. From there two routes migrate a run: the provider said so in words cdx recognises, or — since that wording differs between an enterprise workspace out of credits, a personal plan hitting its 5-hour window, and Claude's own phrasing — simply that the account's quota is gone and the run failed. cdx run-report names which route applied. Exhausting every account reports failover_exhausted, which a caller can tell apart from the task itself failing.
• Native background delegation (opt-in). With CDX_EXPERIMENTAL_NATIVE_BG=1, cdx run --detach hands the run to Claude Code's own background agents (--bg) instead of cdx's detached launcher. The run keeps its cdx run_id, closes out with its token usage and final answer read from the provider's own session transcript, and the parked session is stopped once its task is done. Off by default for one release while it settles; the in-house launcher is unchanged and is what every run takes otherwise. Codex keeps the in-house path while its cloud/exec-server surfaces are marked experimental.
• Observable runs. cdx run-status, cdx run-tail, cdx run-report, and cdx runs --since follow a run while it happens and after it ends.
• Selection without launching. cdx select --provider ... --require-ready --json answers "which session should this job use?" for orchestrators.
• Escape hatch for unmapped flags. cdx set <name> --extra-args "--add-dir ../shared" passes provider arguments cdx does not model straight through, so a flag your provider shipped last week never forces you to abandon cdx. These arguments are explicitly unvalidated: cdx does not parse or verify them, cdx schema --json lists them under unvalidated rather than describing them, and cdx doctor --check-provider-flags ignores them. They are split into literal argv entries and never passed through a shell.
• Validated contract. cdx schema --json publishes the enums, mutually-exclusive argument groups, and error codes programmatic callers should validate against.

Context handoff

• Shared handoff context. Keep a per-workspace Markdown context, or build one from a source session transcript, and install it into another assistant session before switching providers or accounts.
• Session transcript capture. Every launch is recorded to a local log file via script, giving you a full terminal transcript for each session.

Maintenance

• Launch history. Inspect recent launches with provider, result, duration, working directory, launch settings, and transcript path.
• Aggregate usage. cdx stats totals runs, token usage, and time spent per session over any period.

• Disk usage and cleanup. cdx disk reports CDX_HOME usage, cdx disk profiles --candidates identifies reclaimable profile caches/logs with evidence, and cdx clean profiles ... applies explicit cleanup actions.
• Health and repair. cdx doctor inspects dependencies, permissions, and orphan profiles; cdx repair plans or applies safe fixes.
• Portable bundles. cdx export / cdx import move sessions between machines, with passphrase-encrypted auth data when you ask for it.
• Update prompts. Periodic update checks surface cdx update directly in the cdx, cdx status, and launch output when a newer release is available. When logics-manager is installed, cdx can also suggest logics-manager self-update.
• Logics viewer shortcut. cdx view opens the Logics browser/focus viewer through logics-manager view when the companion CLI is installed. All viewer flags are forwarded: --lan, --lan-rw, --focus <ref>, --read, --port, --host, --refresh-interval, --tls, --tls-cert, --tls-key, --open, --no-open. cdx view --json reports availability and update diagnostics without opening the viewer.

Technical Overview

• Python 3.9+.
• Environment isolation per session:
  • Codex sessions override CODEX_HOME to a dedicated profile directory.
  • Claude sessions override HOME to a dedicated profile directory and disable Claude Code commit co-author attribution by default.
  • Antigravity sessions launch agy with a dedicated HOME for file-based settings; Google account credentials may still be stored in the OS keyring by Antigravity itself.
  • Ollama sessions use the local Ollama server and launch ollama run <model>; set a model with cdx set <name> --model <model>.
  • New Codex sessions seed their auth home from your existing global ~/.codex/auth.json when available, so an already logged-in Codex CLI can be reused without giving up per-session isolation afterward.
• Persistence:
  • Session registry at ~/.cdx/sessions.json (versioned JSON store).
  • Per-session state at ~/.cdx/state/<name>.json.
  • Per-workspace shared context at ~/.cdx/contexts/<workspace-hash>/context.md.
  • Auth and provider data under ~/.cdx/profiles/<name>/.
  • All paths are URL-encoded to support arbitrary session names.
• Status resolution pipeline:
  • Primary source: recorded status fields on the session record.
  • Codex live source: codex app-server JSON-RPC account/rateLimits/read, normalized into 5-hour, weekly, reset, credit, and plan fields.
  • Fallback: status-source scans provider JSONL history files and terminal log transcripts, strips ANSI/OSC sequences, and extracts usage%, 5h remaining%, and week remaining% via pattern matching.
• Status latency controls: use cdx status --cached to skip live probes, cdx status --timeout SECONDS for one command, or CDX_STATUS_TIMEOUT_SECONDS to lower the default Codex live-probe timeout.
• Claude status refreshes are cached briefly by default; pass --refresh to force a live rate-limit probe.
• On Linux, transcript capture uses the util-linux script -c command form.
• If script is unavailable, Codex launch falls back to running without transcript capture.
• On Windows, transcript capture is optional. If no compatible script wrapper is installed, Codex still launches normally without transcript capture.
• Auth probe: synchronous subprocess call to codex login status or claude auth status before any interactive launch.
• Signal forwarding: SIGINT, SIGTERM, and SIGHUP are forwarded to the child process and produce clean exit codes.
• Test stack: pytest, pytest-cov, and ruff via the documented dev dependency extra.

Getting Started

Prerequisites

• Node.js 18+ with npm
• Python 3.9+
• codex, claude, agy, and/or ollama CLI installed and available in your PATH

On Windows, the npm launcher looks for Python in this order: py -3, python, then python3. Make sure at least one of those commands resolves to Python 3.

Install

From npm:

npm install -g cdx-manager

With pipx:

pipx install cdx-manager

With uv:

uv tool install cdx-manager

On Windows with PowerShell:

npm install -g cdx-manager

With the standalone PowerShell installer:

Invoke-WebRequest https://raw.githubusercontent.com/AlexAgo83/cdx-manager/main/install.ps1 -OutFile install.ps1
# Optional: set CDX_SHA256 before running if you have a trusted checksum.
# Set CDX_ALLOW_UNVERIFIED=1 only if you intentionally accept an unverified archive.
powershell -ExecutionPolicy Bypass -File .\install.ps1

With the standalone GitHub installer:

curl -fsSL https://raw.githubusercontent.com/AlexAgo83/cdx-manager/main/install.sh -o install.sh
# Optional: set CDX_SHA256 before running if you have a trusted checksum.
# Set CDX_ALLOW_UNVERIFIED=1 only if you intentionally accept an unverified archive.
sh install.sh

For a specific version:

curl -fsSL https://raw.githubusercontent.com/AlexAgo83/cdx-manager/main/install.sh -o install.sh
CDX_VERSION=v0.8.0 sh install.sh

From source:

git clone <repo>
cd cdx-manager
make install

For local development, install the Python toolchain as well:

python3 -m pip install -e ".[dev]"

From source on Windows:

git clone <repo>
cd cdx-manager
npm install -g .

cdx is now available globally. Changes to the source take effect immediately — no reinstall needed.

To update an installed copy later:

cdx update

To uninstall:

make uninstall

To uninstall on Windows after npm install -g:

npm uninstall -g cdx-manager

Alternatively, for a non-symlinked global source install:

npm install -g .

Security note:

• The standalone installers resolve official release checksums from the tagged GitHub Release asset release-archives.json.
• You can still override verification explicitly through CDX_SHA256.
• If no checksum is available, standalone installers fail closed unless CDX_ALLOW_UNVERIFIED=1 is set; that override prints a prominent warning before continuing.
• Prefer npm, pipx, or uv when you want registry-backed install flows.
• If you use the standalone script, download it first, inspect it, and prefer a release with the official checksum asset attached.

Release maintainer note:

• Before publishing npm or PyPI packages, run npm run release:validate.
• The release tag must match package.json, pyproject.toml, src/cli.py, and VERSION.
• checksums/release-archives.json must include the matching vX.Y.Z entry with both github_tarball_sha256 and github_zip_sha256.
• Publishing a GitHub Release runs the checksum upload workflow: it regenerates checksums/release-archives.json for the release tag and uploads it as the release-archives.json asset with replacement enabled for safe reruns.
• If the asset is missing or stale, repair it with python3 scripts/update_release_checksums.py --tag vX.Y.Z, then gh release upload vX.Y.Z checksums/release-archives.json --clobber.
• Publish packages only after the checksum workflow succeeds and npm run release:validate, npm run lint, and npm test pass.

Environment

By default, cdx stores all data under ~/.cdx/. Override with:

export CDX_HOME=/path/to/custom/dir

Optional runtime knobs:

export CDX_CLAUDE_STATUS_MODEL=claude-haiku-4-5-20251001
export CDX_SCRIPT_BIN=script
export CDX_SCRIPT_ARGS='-q -F {transcript}'

PowerShell equivalents:

$env:CDX_HOME = "C:\cdx-data"
$env:CDX_CLAUDE_STATUS_MODEL = "claude-haiku-4-5-20251001"
$env:CDX_SCRIPT_BIN = "script"
$env:CDX_SCRIPT_ARGS = "-q -F {transcript}"

Command Prompt equivalents:

set CDX_HOME=C:\cdx-data
set CDX_CLAUDE_STATUS_MODEL=claude-haiku-4-5-20251001
set CDX_SCRIPT_BIN=script
set CDX_SCRIPT_ARGS=-q -F {transcript}

Quick Start

# Register a Codex session
cdx add work

# Register a Claude session
cdx add claude personal

# Register an Ollama session
cdx add ollama local --model llama3.2

# List all sessions
cdx

# Launch a session
cdx work

# Check usage across all sessions
cdx status

# Pick the best session using the same priority logic as status
cdx next

# Notify when the next cooling-down assistant is ready
cdx ready

# Inspect CDX_HOME and profile disk usage
cdx disk
cdx disk profiles --candidates

Next-Ready Notifications

Agent notifications

When a session's agent finishes a turn, or stops to wait for you, cdx raises a desktop notification naming the session and the repository:

✓ work1     logics-manager · waiting for you
✓ codex-a   cdx-manager · finished

Off until you ask for it, per session:

cdx set work1 --notify on      # turn it on for one session
cdx set --all --notify on      # turn it on everywhere
cdx set work1 --notify off     # and back off; cdx removes what it installed

Turning it on records the setting. The wiring happens at that session's next launch: cdx writes the hook into the session's own home and says so once. Claude Code honours it straight away. Codex only runs hooks that come from an installed plugin, so cdx generates one per session and installs it — and Codex asks you to approve it once, which cdx leaves to you rather than writing that trust on your behalf.

It is off by default because turning it on writes into the provider's own configuration and, on Codex, puts an approval prompt in front of you. That is not something to do to sessions you already own without being asked.

Per platform: macOS uses osascript, Linux notify-send, Windows a toast, and WSL reaches the Windows notification centre through interop. On a host with no way to deliver one — a headless or SSH Linux session, or WSL with interop disabled — cdx installs no hooks at all rather than asking you to approve something that could show you nothing. Headless cdx run never notifies: its caller already learns of completion from the return value.

Use cdx ready when every useful assistant is cooling down and you want your terminal back. cdx picks the next known reset, registers a native OS notification, and exits immediately. If no future reset is known, it exits successfully and explains that no notification was scheduled.

cdx ready
cdx ready --refresh

Persistent Launch Settings

New sessions start with power=medium and fast=off, so launches are predictable without enabling fast mode. Set or override only the values you want to pin:

cdx set work --power medium --permission full --fast off
cdx set personal --power low --permission review
cdx set --sessions all --permission auto
cdx set --provider ollama --model llama3.2
cdx set work --priority 80
cdx set work --rtk on
cdx set work --logics off
cdx power all low
cdx perm provider:claude review
cdx model provider:ollama llama3.2
cdx config work
cdx configs

Those values are stored on the session and reapplied every time you run cdx work. Remove overrides to return to provider-native defaults:

cdx unset work --power
cdx unset --sessions work,personal --fast
cdx unset --provider claude --permission
cdx unset work --priority
cdx unset work --rtk
cdx unset work --logics
cdx unset work --all
cdx power all default
cdx model provider:ollama default

--model maps to Codex --model, Claude --model, and Ollama ollama run <model>. --power maps to Codex model_reasoning_effort and Claude --effort; supported values are minimal, low, medium, high, and xhigh. --permission maps to provider-native permission flags. Codex Fast mode is separate from reasoning effort: --fast on opts a Codex session into the Codex Fast service tier, while new sessions, --fast off, and default launches force the non-Fast flex tier. Use --power low when you want low reasoning effort without enabling Codex Fast credits. Existing legacy sessions that stored fast=on before this split continue to behave as low effort unless the user explicitly sets --fast on again. --priority is a 0..100 selector preference used as a tie-breaker after readiness and availability. --rtk on injects a launch instruction that encourages assistants to use RTK (rtk <command>) for noisy terminal commands when RTK is available, while keeping raw commands for exact output. Logics guidance is auto-enabled when logics-manager is available; use --logics off to disable that guidance for a session, or --logics on to pin it explicitly.

Launch History

Every interactive cdx <name> launch is recorded under CDX_HOME, including success/failure, duration, cwd, launch settings, and transcript path.

cdx history
cdx history work
cdx history work --limit 5
cdx history work --json
cdx history --summary
cdx history --summary --since 7d
cdx history --summary --from 2026-05-01 --to 2026-05-28

cdx history --summary aggregates total time per assistant. Add --since, --from, or --to to focus on a period.

All Commands

JSON Output

cdx-manager can be consumed by other apps through its CLI JSON contract.

Commands with machine-readable output:

• cdx --json
• cdx status --json
• cdx status <name> --json
• cdx add ... --json
• cdx cp ... --json
• cdx ren ... --json
• cdx label ... --json
• cdx rmv ... --json
• cdx clean ... --json
• cdx disk ... --json
• cdx reset ... --json
• cdx export ... --json
• cdx import ... --json
• cdx login ... --json
• cdx logout ... --json
• cdx disable ... --json
• cdx enable ... --json
• cdx context ... --json
• cdx memory ... --json
• cdx handoff ... --json
• cdx history ... --json
• cdx stats ... --json
• cdx last --json
• cdx doctor --json
• cdx repair --json
• cdx view --json
• cdx update --json
• cdx ready --json
• cdx select ... --json
• cdx run ... --json

Success payloads follow a shared envelope:

{
  "schema_version": 1,
  "ok": true,
  "action": "add",
  "message": "Created session work (codex)",
  "warnings": [],
  "session": {
    "name": "work"
  }
}

Most commands use a shared stderr JSON envelope for errors whenever --json is present:

{
  "schema_version": 1,
  "ok": false,
  "error": {
    "code": "invalid_usage",
    "message": "Usage: cdx status [--json] [--refresh|--cached] [--timeout SECONDS] | ...",
    "exit_code": 1
  }
}

status --json and similar commands also use the same envelope and place non-fatal issues in warnings instead of mixing plain-text diagnostics into stderr. cdx run --json is the exception: it always writes one final JSON payload to stdout, including cdx-side and provider-start errors, so supervisors can parse a single result stream while provider stdout and stderr are captured to files.

This makes cdx-manager usable from editor plugins, scripts, and desktop apps without scraping human-readable terminal output.

Headless Runs

cdx run is designed for supervisors such as Orchestia. In --json mode, stdout contains only the final JSON payload; provider stdout and stderr are captured to files.

cdx run codex-work \
  --cwd /path/to/workspace \
  --prompt-file task_prompt.md \
  --model gpt-5.3-codex \
  --reasoning-effort low \
  --permission workspace-write \
  --timeout-seconds 1800 \
  --json

Use provider-based auto-selection when the caller wants cdx-manager to pick the account:

cdx run \
  --provider codex \
  --cwd /path/to/workspace \
  --prompt "Summarize the repo status." \
  --reasoning-effort low \
  --json

The result includes launcher: "cdx", run_id, selected session, provider, exit_code, duration_seconds, absolute transcript_path, stdout_path, stderr_path, and normalized usage token fields. Codex headless runs use codex exec --json; Claude headless runs use claude --print --output-format json. Token counts are null when the provider does not expose a supported JSON or JSONL usage shape.

Known headless token usage is persisted in launch history and can be aggregated later:

cdx stats --since 7d --json
cdx stats work

cdx select exposes the same session selection logic directly:

cdx select --provider codex --min-reasoning-effort low --require-ready --json

How cdx picks a session

cdx select, cdx run --provider, cdx next, the cdx status recommendation, and cdx ready all use one ranking. They used to use two that disagreed, so the answer depended on which command you asked.

Candidate filters exclude sessions before any ordering happens: disabled sessions, logged-out sessions (they cannot serve work), sessions of another provider when one is requested, sessions below a requested --min-reasoning-effort, and — when readiness is required — sessions with no availability left.

Ordering compares the remaining candidates factor by factor:

1. Usability — usable now, then blocked with a known upcoming reset, then reset known, then nothing known.
2. Priority — the --priority you set with cdx set <name> --priority 0..100, highest first.
3. Then, for a session you can use now: credits (sessions without credits first, so included quota is spent before paid credits), availability, reset time. For a session you cannot use now the order is reset time first, then credits and availability — when you cannot use it, what matters is when it comes back.
4. Reasoning effort — the lowest configured effort that still clears any requested floor, leaving stronger sessions free for work that needs them.
5. Session name, so the order is deterministic.

--priority ranks sessions within a usability class, never across one. A high-priority session with no quota left does not outrank a usable one: priority expresses which usable session to prefer, not a way to route work somewhere it will fail.

cdx select --json reports selection_policy (built from the ranking itself, so it cannot describe an order the code does not apply) and deciding_factor/reason, naming the factor that actually separated the winner from the runner-up for that call — or reporting that it was the only candidate. Do not treat selection_policy as a stable identifier; it changes when the ranking changes, which is the point.

cdx run --provider selects from cached status by default. Pass --refresh to fetch status first. When a session is auto-selected without any recorded availability, the run payload carries a session_selected_without_status warning — that is different from a session known to be low, and on a freshly imported or long-idle set of sessions it is the ordinary case.

Verifying provider flag mappings

cdx translates each --permission value into concrete provider CLI flags. Those mappings are declarations about someone else's CLI, and a provider can drop or rename a flag without cdx noticing — which is exactly what happened with an --experimental-yolo flag mapped for ollama that the ollama CLI never had.

cdx doctor --check-provider-flags verifies them against the installed provider CLIs:

cdx doctor --check-provider-flags --json

For each configured provider it reports, per permission, whether every mapped flag is accepted. A rejected flag is a FAIL naming the provider, the permission, and the flag. A provider whose CLI is not installed, or whose help cannot be read, is a WARN — never an OK: "could not verify" is precisely the state a stale mapping hides in. A provider that maps no flags at all, such as ollama, reports OK with an empty mapping, so having nothing to check is distinguishable from having failed to check.

It is opt-in because it costs one provider CLI invocation per configured provider. When it is not requested, the default cdx doctor report says so with a provider_permission_flags_unchecked warning rather than omitting it, so a green report never implies the mappings were verified.

Programmatic Callers

cdx run blocks until the provider finishes, which suits a supervisor that can wait. Agents, MCP servers, and watchdogs usually cannot, so the surface below exists specifically for them: launch without waiting, watch progress, react to failures by code, and discover valid values instead of hard-coding them.

Launch without waiting. --detach returns as soon as the run is registered, with the run_id already assigned:

cdx run work --cwd /path/to/repo --prompt-file task.md --detach --json

The payload carries detached: true, run_id, pid, and the artifact paths — but no exit_code, duration_seconds, or usage, since the run has not finished. The detached process is put in its own session, so it survives the launcher exiting, including when cdx was invoked over an SSH command that returns immediately. It transitions its own registry entry to a terminal status, so cdx runs and cdx run-status stay accurate with nobody supervising.

Authentication is still resolved before the launch returns, so a login problem surfaces synchronously rather than silently in the background.

Watch a run in progress. cdx run-status reports only running/succeeded/failed; run-tail shows what the run is actually doing:

cdx run-tail <run_id> --lines 50 --json

It returns the last lines of the run's own stdout_path plus its current status, whether the run is still going or already finished. Lines are raw provider output, not summarized. Undecodable bytes come back with replacement characters rather than failing the call.

Pipe an untrusted prompt. --prompt-file - reads the prompt from standard input, so arbitrary text never has to reach a command line or a temporary file the caller must clean up:

echo "$UNTRUSTED_PROMPT" | cdx run work --cwd /path/to/repo --prompt-file - --json

It fails immediately if standard input is a terminal, rather than blocking at a silent prompt.

Poll for completions with a cursor. cdx runs --since returns every run that completed after the cursor, using the same cursor forms as cdx history --since:

cdx runs --since 15m --json
cdx runs --since 2026-08-07T09:00:00Z --json

A cursor is bounded by time, not by row count: --limit is ignored when --since is given (and says so in warnings). That is deliberate — capping a cursor query by row count would silently drop completions a caller had not yet seen, which is exactly the miss the cursor removes. Runs still in flight are not returned; they appear once they complete. Callers do not need to keep their own set of already-reported run ids.

React to failures by code, not by message. Argument failures report a specific error.code and name the offending arguments as data:

{
  "ok": false,
  "error": {
    "code": "mutually_exclusive_arguments",
    "message": "cdx run: cannot specify both a session name and --provider.",
    "arguments": ["session", "--provider"],
    "allowed_values": null
  }
}

Never branch on error.message; it is written for a human reading a terminal and may be reworded.

Discover valid values. cdx schema --json publishes the enums and argument constraints the parser itself validates against:

cdx schema --json

It returns accepted values for permission (with its aliases and canonical forms), reasoning_effort/power, kind, and provider, plus the declared mutually-exclusive argument groups and the error codes above. Validate against this rather than copying the lists into your own code — a hand-maintained copy drifts, and cdx will not tell you when it does.

Surface run warnings. The run payload's warnings list reports degradations a zero exit code would otherwise hide, on the successful path as much as the failing one:

{
  "ok": true,
  "exit_code": 0,
  "warnings": [
    {
      "code": "network_disabled_by_permission",
      "message": "codex runs at permission 'review' inside a sandbox with no network access, ...",
      "provider": "codex",
      "permission": "review"
    }
  ]
}

Codex ties network access to its sandbox, so any permission below full leaves the run unable to resolve DNS — gh, curl, and package installs fail inside an otherwise successful run. Show these warnings to whoever reads your output rather than discarding them; a run that succeeds while quietly doing less is the case they exist for.

Backup And Restore

You can move sessions between machines with portable bundles:

cdx export backup.cdx
cdx import backup.cdx

To migrate auth and avoid logging in again, include auth data in an encrypted bundle:

export CDX_BUNDLE_PASSPHRASE='choose-a-strong-passphrase'
cdx export backup-auth.cdx --include-auth --passphrase-env CDX_BUNDLE_PASSPHRASE
cdx import backup-auth.cdx --passphrase-env CDX_BUNDLE_PASSPHRASE

Notes:

• --include-auth is encrypted, requires a passphrase, and exports only provider credential files rather than full profile caches, logs, or installed plugins.
• Without --passphrase-env, cdx prompts in an interactive terminal. Non-interactive callers can pass --passphrase-stdin to read the passphrase from stdin (one line) instead of exposing it in the child environment.
• --sessions work,perso exports or imports only a subset.
• --force allows overwriting existing destination sessions during import or replacing an existing bundle file during export. Import preserves detected local plugins/ state when the bundle does not provide it, and refuses to overwrite existing sessions from a bundle without auth payloads unless --allow-authless-force is also passed.
• Auth bundles contain credentials. Treat them like secrets and delete them after transfer.

Available Scripts

• npm test: run the Python test suite with pytest
• npm run test:py: run pytest through the portable launcher
• npm run test:coverage: run pytest with terminal coverage reporting for src/
• npm run lint: check project guidance, the Node launcher, ruff, and byte-compile the Python sources, scripts, and tests
• npm run release:validate: verify version alignment and required GitHub release archive checksum metadata before publication
• npm run link: link cdx globally for local development (npm link)
• npm run unlink: remove the global link

Windows Support

• Supported install paths on Windows:
  • npm install -g cdx-manager
  • pipx install cdx-manager
  • uv tool install cdx-manager
  • install.ps1
• The npm launcher resolves Python via py -3, python, then python3, so a global npm install works even when python3.exe is missing.
• install.sh is Unix-only.
• make install and make uninstall are Unix-oriented convenience commands, not the default Windows path.
• cdx isolates Claude sessions on Windows by setting HOME, USERPROFILE, HOMEDRIVE, and HOMEPATH.
• Desktop notifications use PowerShell on Windows.
• Codex transcript capture is optional on Windows:
  • if a compatible script command is available and exposed via CDX_SCRIPT_BIN, cdx uses it
  • otherwise Codex launches without transcript capture and the session still works normally
• cdx doctor reports the transcript-capture fallback explicitly so missing script on Windows is visible without being treated as a hard failure.
• cdx doctor --json includes detected codex --version and claude --version output when available, plus conservative capability hints for recent provider memory, model, and headless diagnostic surfaces. These hints are informational; cdx does not edit provider-private memory stores.

Project Structure

bin/
  cdx.js                # Node launcher used by npm
  python-runner.js      # Shared Python resolver and process wrapper
  cdx                   # Python entrypoint invoked by the launcher

src/
  cli.py                # Top-level command router
  cli_commands.py       # Command handlers and argument handling
  cli_render.py         # Terminal formatting, tables, colors, and errors
  backup_bundle.py      # Portable session bundle encoding/decoding + auth encryption
  status_view.py        # Status table/detail rendering and priority ranking
  provider_runtime.py   # Provider launch/auth commands, transcripts, signals
  claude_refresh.py     # Claude usage refresh orchestration
  session_service.py    # Session lifecycle: create, copy, rename, launch, remove, status
                        # resolution, auth state management
  session_store.py      # JSON persistence layer: sessions.json + per-session
                        # state files
  status_source.py      # Status artifact discovery: scans JSONL history files
                        # and terminal log transcripts, strips ANSI sequences,
                        # extracts usage metrics via pattern matching
  config.py             # CDX_HOME resolution (env override or ~/.cdx)
  errors.py             # CdxError with optional exit code
  __init__.py           # Public Python exports

test/
  test_cli_py.py            # CLI command dispatch tests
  test_session_service_py.py  # Session service unit tests

Data Layout

All session data lives under CDX_HOME (default: ~/.cdx/):

~/.cdx/
  sessions.json             # Session registry (versioned, all sessions)
  state/
    <encoded-name>.json     # Per-session rehydration state
    launch_history.jsonl    # Append-only launch history
  contexts/
    <workspace-hash>/context.md     # Current-workspace shared context / memory
    global/context.md               # Global cdx memory
    projects/<encoded-name>/context.md  # Named project memory from cdx memory --project NAME
  profiles/
    <encoded-name>/         # Codex session: CODEX_HOME points here
      log/
        cdx-session.log     # Terminal transcript (written by script(1))
    <encoded-name>/
      claude-home/          # Claude session: HOME points here
        log/
          cdx-session.log

Session names are URL-encoded when used as directory or file names. CLI command names such as add, disk, status, and login are reserved and cannot be used as new session names.

cdx memory is the user-facing interface for explicit memory. --project NAME uses a stable free-form project name; --project /path/to/repo resolves to the same workspace-hash memory that repo would use as the current workspace. append adds the text as-is with a clean newline boundary; it does not add dates or headings. list reports global memory, named project memories, and the current workspace memory when present; it does not scan your filesystem for repos.

Provider-private memory files, such as Codex SQLite memory databases or Claude Code project memory files, are not edited by cdx memory.

Disk maintenance

Use cdx disk for a total CDX_HOME measurement, and cdx disk profiles for the profiles/ total plus a per-profile breakdown.

In an interactive terminal, long scans report their current stage and per-profile progress on stderr. JSON and redirected output remain free of progress messages.

Text output uses aligned summary tables with profile size, share of total space, and reclaimable space. Candidate details are grouped by profile with separate size, type, risk, and evidence columns.

cdx disk
cdx disk profiles

To find cleanup candidates without deleting anything:

cdx disk profiles --candidates
cdx disk profiles --candidates --json

Candidate evidence includes profile name, path, size, risk, reason, and metadata such as modified time or old-log file counts. The current cleanup candidates are intentionally limited to:

• temporary marketplace/plugin staging caches under profiles/*/.tmp/
• .log files older than the requested age inside profile log/ directories

cdx also runs a passive disk cleanup check at most once per day on advisory commands such as cdx and cdx status. It writes the check timestamp and latest measurements to ~/.cdx/state/disk-cleanup-check.json. A warning is shown only when at least one threshold is met:

• CDX_HOME uses 10 GB or more
• cleanup candidates can reclaim 1 GB or more
• temporary cache candidates can reclaim 500 MB or more
• old log candidates can reclaim 500 MB or more
• one profile is 2 GB or more, or one profile is at least 25% of profiles/ when profiles/ is already 1 GB or more

The passive warning is read-only and points to cdx disk profiles --candidates; it never deletes files.

Apply cleanup only with explicit actions:

cdx clean profiles --tmp
cdx clean profiles --old-logs 30d

--tmp removes temporary marketplace/plugin clone/backup staging directories. --old-logs 30d removes only .log files older than 30 days. The commands do not remove auth.json, config.toml, sessions/, SQLite state, installed plugins/, skills/, or credentials.

All cdx clean actions require interactive confirmation before any deletion or log truncation begins. Non-interactive scripts must pass --yes explicitly.

The cleanup action is selected only when --tmp or --old-logs is present. This keeps cdx clean profiles available for an existing session named profiles. If any candidate or old log cannot be removed, cleanup stops with a non-zero error that identifies the affected path; it does not report the operation as successful.

Banked Codex resets

Eligible Codex accounts expose manually redeemable rate-limit resets through the Codex app-server. cdx status shows their available count in the RESETS column; cdx status <name> also shows the nearest known expiration, and JSON status includes the available reset details supplied by Codex.

Activation is always explicit:

cdx reset <name>
cdx reset <name> --yes --json

Without --yes, an interactive confirmation is required. cdx sends an idempotency key, consumes at most one available reset, and refreshes the session status after a successful activation. It never activates a reset from cdx status or any background check.

Troubleshooting

• cdx <name> fails with "not authenticated" — run cdx login <name> first.
• One of two Codex accounts keeps asking for login — run cdx doctor --json and inspect each Codex session's codex_auth_file, codex_live_auth, and codex_stale_auth_logs checks. If recent logs mention expired auth, repair only that isolated profile with cdx login <name>. Codex Business profiles can share the same tokens.account_id; cdx doctor reports codex_shared_account_id as a workspace-level clue, not proof that two profiles use the same user. Newly created sessions seed from the current global ~/.codex/auth.json when one exists. For two separate Codex accounts, create or repair each session by running cdx login <name> for that session; cdx login does not log out first, so use cdx logout <name> explicitly only when you want to clear that isolated profile.
• cdx says no compatible Python 3 interpreter was found — install Python 3 and make py -3, python, or python3 available on PATH.
• cdx add succeeds but the session does not appear — check that CDX_HOME is consistent between calls; a mismatch creates two separate registries.
• Status shows n/a for all fields — the Codex app-server rate-limit probe may be unavailable, the session may not be authenticated, and no legacy transcript/history status has been captured yet.
• cdx rmv says "Removal requires confirmation in an interactive terminal" — pass --force to bypass the prompt in non-interactive environments (scripts, CI).
• cdx login hangs — the provider's login flow requires a browser or device code. Follow the on-screen instructions in the terminal that opened.
• make install says npm link is not found — ensure Node.js and npm are installed and in your PATH.
• On Windows, doctor warns that script is missing — this is expected on many setups. Codex still launches, but transcript capture stays disabled unless you point CDX_SCRIPT_BIN to a compatible wrapper.

Contributing

Contribution guidelines are available in CONTRIBUTING.md.

License

This project is licensed under the MIT License. See LICENSE.